INTRODUCTION: The study of cranio-facial morphological variations is useful for describing and classifying skeletal relations. It considers univariate and multivariate analysis using variables such as distances, angles and reference planes. These methods provide a partial and localized description of shape, while the methods based on geometric morphometrics (GM) in two or three dimensions (2D, 3D) allow a more sensible examination of variables and detailed comprehension of associations. The GM constitutes a new analysis tool of form variations representing a notable advance regarding other morphometric methods. OBJECTIVE: to characterize the mandibular morphology in I, II, III skeletal relations using 2D geometric morphometry. METHODOLOGY: from an analytical cross-sectional study, 60 cranial lateral images (30 males, 30 females) with I, II and III skeletal relations were taken from Colombians aged between 17 and 27 years. It yielded coordinates of 14 facial polygon landmarks and 13 mandibular, for the description of both structures. Intra-examiner error was determined with ANOVA-II and atypical data detection with bi-variate bagplot distribution. A generalized analysis of Procrustes (GPA) on landmarks in order to obtain the centroid size (CS) and the uniform and non-uniform components of conformation (PW). Size variation of structures was evaluated with the Kruskal-Wallis applied to the CS and conformational differences significance between structures with multivariate variance analysis (Anova of Procrustes). Facial polygon and mandible variations identification was obtained with a principal component analysis (PCA) and canonical variable analysis (CVA), description and visualization of these variations was done with a relative deformation analysis. A discriminant analysis from the obtained canonical variables was also carried out. Lastly, a K-medians clustering (k=4) analysis was carried out, using the facial and mandible polygon partial deformations. RESULTS: The intra-observer error was less than 28.5% at the prosthion point. On the PCA, components 1 and 2 support and 80% variance. The CVA showed separation between groups, with class II and class III in female facial polygon the most dissimilar. The discriminant analysis showed a high classification capacity. CONCLUSIONS: The conformation study using 2D GM showed an adequate sensibility for capturing and analyzing morphological variations of the assessed structures. Conformational variation of the mandible was distributed along several components, which must be evaluated in a synergic manner accumulating more and better elements of judgement for clinical decisions.